%reload_ext autotime
import pandas as pd
import requests
from pprint import pprint
import json
import torch
from PIL import Image
from transformers import MllamaForConditionalGeneration, AutoProcessor
from tqdm.auto import tqdm

pd.options.plotting.backend = "plotly"
pd.set_option("display.max_columns", None)
pd.set_option("display.max_colwidth", 100)

✔️ 4.79 s (2024-12-12T10:23:55/2024-12-12T10:24:00)

2024-12-12 10:23:58.385012: I tensorflow/core/util/port.cc:153] oneDNN custom operations are on. You may see slightly different numerical results due to floating-point round-off errors from different computation orders. To turn them off, set the environment variable `TF_ENABLE_ONEDNN_OPTS=0`.
2024-12-12 10:23:58.403285: E external/local_xla/xla/stream_executor/cuda/cuda_fft.cc:485] Unable to register cuFFT factory: Attempting to register factory for plugin cuFFT when one has already been registered
2024-12-12 10:23:58.423118: E external/local_xla/xla/stream_executor/cuda/cuda_dnn.cc:8454] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered
2024-12-12 10:23:58.429643: E external/local_xla/xla/stream_executor/cuda/cuda_blas.cc:1452] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered
2024-12-12 10:23:58.448438: I tensorflow/core/platform/cpu_feature_guard.cc:210] This TensorFlow binary is optimized to use available CPU instructions in performance-critical operations.
To enable the following instructions: AVX2 AVX512F AVX512_VNNI FMA, in other operations, rebuild TensorFlow with the appropriate compiler flags.
2024-12-12 10:23:59.284953: W tensorflow/compiler/tf2tensorrt/utils/py_utils.cc:38] TF-TRT Warning: Could not find TensorRT

df = pd.read_csv("results.csv").drop_duplicates(subset="panoid")
df

✔️ 37 ms (2024-12-12T10:24:00/2024-12-12T10:24:00)

# Loading this model needs about 22.69GB of GPU memory
model_id = "meta-llama/Llama-3.2-11B-Vision-Instruct"

model = MllamaForConditionalGeneration.from_pretrained(
    model_id,
    torch_dtype=torch.bfloat16,
    device_map="auto",
)
processor = AutoProcessor.from_pretrained(model_id)

✔️ 13.2 s (2024-12-12T10:24:00/2024-12-12T10:24:13)

The model weights are not tied. Please use the `tie_weights` method before using the `infer_auto_device` function.

Loading checkpoint shards:   0%|          | 0/5 [00:00<?, ?it/s]

for row in tqdm(df.sample(10).itertuples(index=False)):
    panoid = row.panoid
    image = Image.open(f"panoramas/{panoid}.jpg")
    display(image)
    messages = [
        {
            "role": "user",
            "content": [
                {"type": "text", "text": """
                    This image is a panorama from Google Street View.
                    From the image, extract the following information, in JSON format:
                    green: Percentage of the image that is green space (e.g. parks, gardens, trees, grass etc.). A number from 0-100.
                    environment: Classify the nature of the environment in this image. Built up/green/residential/shops/cafes?. A string.
                    water: If you see any streams/ponds/rivers/ocean in the image, estimate the distance to the water in meters. A number. If there is no water, return 0.
                    obscured: Proportion of view obscured by buildings (how much of total line of sight is blocked by buildings in close proximity). A number from 0-100.
                    people: the number of people you see in the image
                    cars: the number of cars you see in the image
                    bikes: the number of bikes you see in the image

                    Do not include comments in your JSON response. Only respond with the JSON object. Make sure the JSON is valid.
                """},
                {"type": "image"},
            ]
        }
    ]
    input_text = processor.apply_chat_template(messages, add_generation_prompt=True)
    inputs = processor(
        image,
        input_text,
        add_special_tokens=False,
        return_tensors="pt"
    ).to(model.device)

    for retry in range(3):
        output = model.generate(**inputs, max_new_tokens=5000)
        result = processor.decode(output[0])
        result = result[result.rindex("<|end_header_id|>") + len("<|end_header_id|>"):].strip().replace("<|eot_id|>", "")
        print("Output:")
        try:
            result = json.loads(result)
            pprint(result)
            print("\n")
            break
        except json.JSONDecodeError:
            print(f"Unable to parse: {result}")

✔️ 26.6 s (2024-12-12T10:24:13/2024-12-12T10:24:40)

0it [00:00, ?it/s]

Output:
{'bikes': 0,
 'cars': 2,
 'environment': 'shops',
 'green': 5,
 'obscured': 25,
 'people': 3,
 'water': 0}

Output:
{'bikes': 0,
 'cars': 3,
 'environment': 'residential',
 'green': 60,
 'obscured': 30,
 'people': 0,
 'water': 0}

Output:
{'bikes': 0,
 'cars': 2,
 'environment': 'residential',
 'green': 54,
 'obscured': 55,
 'people': 0,
 'water': 0}

Output:
{'bikes': 0,
 'cars': 3,
 'environment': 'residential',
 'green': 40,
 'obscured': 20,
 'people': 0,
 'water': 0}

Output:
{'bikes': 0,
 'cars': 10,
 'environment': 'residential',
 'green': 30,
 'obscured': 40,
 'people': 0,
 'water': 0}

Output:
{'bikes': 0,
 'cars': 3,
 'environment': 'residential',
 'green': 60,
 'obscured': 50,
 'people': 0,
 'water': 0}

Output:
{'bikes': 0,
 'cars': 1,
 'environment': 'residential',
 'green': 30,
 'obscured': 50,
 'people': 0,
 'water': 0}

Output:
{'bikes': 0,
 'cars': 3,
 'environment': 'residential',
 'green': 43,
 'obscured': 0,
 'people': 0,
 'water': 0}

results = []
for row in tqdm(df.itertuples(index=False), total=len(df)):
    panoid = row.panoid
    image = Image.open(f"panoramas/{panoid}.jpg")
    #display(image)
    messages = [
        {
            "role": "user",
            "content": [
                {"type": "text", "text": """
                    This image is a panorama from Google Street View.
                    From the image, extract the following information, in JSON format:
                    green: Percentage of the image that is green space (e.g. parks, gardens, trees, grass etc.). A number from 0-100.
                    environment: Classify the nature of the environment in this image. Built up/green/residential/shops/cafes?. A string.
                    water: If you see any streams/ponds/rivers/ocean in the image, estimate the distance to the water in meters. A number. If there is no water, return 0.
                    obscured: Proportion of view obscured by buildings (how much of total line of sight is blocked by buildings in close proximity). A number from 0-100.
                    people: the number of people you see in the image
                    cars: the number of cars you see in the image
                    bikes: the number of bikes you see in the image

                    Do not include comments in your JSON response. Only respond with the JSON object. Make sure the JSON is valid.
                """},
                {"type": "image"},
            ]
        }
    ]
    input_text = processor.apply_chat_template(messages, add_generation_prompt=True)
    inputs = processor(
        image,
        input_text,
        add_special_tokens=False,
        return_tensors="pt"
    ).to(model.device)

    for retry in range(3):
        output = model.generate(**inputs, max_new_tokens=5000)
        result = processor.decode(output[0])
        result = result[result.rindex("<|end_header_id|>") + len("<|end_header_id|>"):].strip().replace("<|eot_id|>", "")
        #print("Output:")
        try:
            result = json.loads(result)
            #pprint(result)
            row = row._asdict()
            row.update(result)
            results.append(row)
            print("\n")
            break
        except json.JSONDecodeError:
            print(f"Unable to parse: {result}")

results = pd.DataFrame(results)
results.to_csv("LLM_results.csv", index=False)
results

⌛ 2.03 µs (2024-12-12T10:25:02)

  0%|          | 0/592 [00:00<?, ?it/s]

	Index	pid	n	time	anxiousness	latitude	longitude	geometry	panoid	panolat	panolon	panodate
0	0	P20001	1	2023-04-25T02:51:42Z	0	-36.924795	174.738044	POINT (174.7380435 -36.92479483)	IvrcS0W1RlFAlnci-p39XA	-36.924665	174.737914	2012-04
10	10	P20001	11	2023-04-24T00:42:25Z	0	-36.924837	174.737948	POINT (174.7379477 -36.92483659)	QEpZV7bnO2mBfp0weMUKEg	-36.924730	174.737826	2012-04
13	13	P20006	1	2023-06-03T02:45:55Z	3	-36.892203	174.740125	POINT (174.7401253 -36.89220256)	omb98QNjTPWi0uUfMsmYeg	-36.892621	174.739961	2024-05
14	15	P20009	2	2023-05-17T04:54:48Z	3	-36.923191	174.748620	POINT (174.7486203 -36.92319093)	E7B5AV3DQ1rYWDClVRo8Zg	-36.923194	174.748831	2024-05
17	19	P20009	6	2023-05-19T22:28:51Z	1	-36.923260	174.748655	POINT (174.748655 -36.92325959)	KCTcsxYCIm41XdzkYEYUQw	-36.923286	174.748840	2024-05
...	...	...	...	...	...	...	...	...	...	...	...	...
1314	1421	P20693	2	2024-05-02T03:43:23Z	3	-36.897778	174.721580	POINT (174.7215796 -36.89777786)	Uzuqd6oSo-EjCVuRP2Os0Q	-36.897742	174.721877	2022-06
1317	1425	P20693	6	2024-05-05T03:00:22Z	2	-36.969426	174.790602	POINT (174.7906024 -36.96942642)	4OskePS4Ilz12JhsP-1ujg	-36.969164	174.790848	2022-08
1318	1426	P20721	1	2024-05-05T02:00:52Z	1	-36.893455	174.728262	POINT (174.728262 -36.89345532)	CfRtPfDMNhfXHTNvMwnYRw	-36.893394	174.728062	2024-06
1320	1428	P20721	3	2024-05-05T23:06:27Z	2	-36.845252	174.759951	POINT (174.7599508 -36.8452515)	AF1QipN2FD2eYEmK8bRpEgoM7fFl5-nUstwWujnRj0gv	-36.845292	174.759939	2022-06-24
1321	1429	P20721	4	2024-05-06T07:04:57Z	0	-36.845165	174.759885	POINT (174.7598849 -36.84516487)	AF1QipNj6yheGtCvR6Gk2Svq_lG_fuaGPjehPV8kouy8	-36.845177	174.759792	2022-06-24